软件产品质量度量培训课件.pptVIP

* 5.测试过的系统数量 所支持的不同语言系统的总数与测试过的语言系统数量 所支持系统的总数与测试过的系统数量 Windows 2000 (SPx) Windows XP (SPx) Windows 2003 Server (SPx) Tablet PC 新的系统平台 * 6.自动化测试度量 测试的可自动化程度 能自动化的和实现自动化的比率 运行通过的自动化脚本比率 不同产品开发阶段该比率变化 * 7.Code Coverage（代码覆盖） 代码覆盖度量定义和目的 代码覆盖种类 代码覆盖的有效使用 开发人员：单元测试 (unit testing) 测试人员：系统测试(system testing)和自动化测试 * 代码覆盖是什么？ 动态白盒测试评价技术 已经执行（测试）了什么（what has been executed） 没有执行的（测试）有什么 what has not been executed and still remains to be tested. 需要有源代码 内部辅助工具 * Product Quality can be indicated by: MTTF (from monkeys, overnight stress tests, or end users), Tester MTTF can be problematic, but TM watched bugs per test hours drop by a factor of four before ship: it had a long stabilization period, which helped. NT’s stress metric is a good example of a useful MTTF measure, they dropped direct MTTF measures (from 1000+ users measured for 8 days) because the stress data (from 3-400 machines overnight) can be collected more quickly and it tracked the MTTF metric closely: they have a metric 95% pass rate required before Subjective quality (as determined by the responsible tester or lead), Code churn --excessive churn can indicate bugs to be found (filtered for new features). \\toolsvr\tools\testing\tracking\churn file:\\itg1\contrib\churn3 \\toolsvr\tools\testingtracking\scvt Currently no one has build a code Churn extractor for VSS, anyone interested can check the Testinfo Public Folder for ideas on how to do it, and email jamesti if you do (so others can use it). [Sean McDowell] A useful predictor of how quickly the bug rate will converge to zero is to graph number of lines of code added, changed, and deleted per subsystem/module. It’s also helpful at times to observe the number of new bugs in a module that has a lot of turmoil. If there arent very many this can be an indicator of poor test coverage. KLOC useful as well (growth absolute). MSKLOC file:\\toolsvr\tools\toolbox\mskloc NCKLOC file:\\itg1\contrib\nckloc Number of builds since last IDW. IDW (internal development